Computer Vision Approaches for Extracting Fire-Safety Information from Service Drawings
(2026) EITM01 20252Department of Electrical and Information Technology
- Abstract
- Fire-safety documentation is an essential part of regular inspections during the building operations and maintenance phase. As help, service drawings exist to detail the positions of various fire-safety devices and their addresses, as well as fire zone coverage throughout the floors and sectors of a building. Automatically extracting this information would be immensely helpful towards, for example, maintaining up-to-date documentation or integration onto digital systems. To that end, this thesis adapts multiple computer vision techniques in order to extract pertinent information: (1) a Keypoint R-CNN model is trained on a custom dataset to detect the symbols of fire devices along with their installation positions, (2) OCR is used to parse... (More)
- Fire-safety documentation is an essential part of regular inspections during the building operations and maintenance phase. As help, service drawings exist to detail the positions of various fire-safety devices and their addresses, as well as fire zone coverage throughout the floors and sectors of a building. Automatically extracting this information would be immensely helpful towards, for example, maintaining up-to-date documentation or integration onto digital systems. To that end, this thesis adapts multiple computer vision techniques in order to extract pertinent information: (1) a Keypoint R-CNN model is trained on a custom dataset to detect the symbols of fire devices along with their installation positions, (2) OCR is used to parse addresses and (3) a region merging procedure is implemented to segment fire zones. Additionally, symbols are also assigned to their addresses using the Hungarian algorithm, with Euclidean distance as base cost.
For the symbol detection, we find that crucial model settings had to be changed in order to tailor the model for the domain and achieve satisfactory performance. High accuracy is achieved despite major class-imbalances in the dataset, but the model fails in all cases to detect the true installation positions of devices, whenever these are indicated by an auxiliary line. Assigning symbols and addresses to each other based on distance works well in most cases, while having a few edge cases that are impossible to amend with our method alone, due to inherent quirks in the service drawings. Using region merging for the fire zone segmentation gives us high recall, but low precision. The low precision is mainly due to the oversegmentation step of region merging, and partly due to the existence of hard-to-ignore false positives in most service drawings. Overall, the results across all tasks are promising while still leaving plenty of room for improvement, or alternative approaches, in further work. (Less) - Popular Abstract
- Fire-safety is a critical aspect of buildings where, in extreme cases, human lives are at stake. As required by law, documentation exists to record the locations of various fire-safety devices (e.g. smoke detectors) in order to support proper maintenance and regulatory compliance. This thesis mainly explores, to great success, the use of artificial intelligence to automatically extracting information from such documents.
Automating this process could reduce time-consuming documentation work and make it easier to keep fire-safety systems up to date—especially for larger facilities such as hospitals, universities, or industrial sites where hundreds of documents may need to be maintained and regularly revised. Although these documents are... (More) - Fire-safety is a critical aspect of buildings where, in extreme cases, human lives are at stake. As required by law, documentation exists to record the locations of various fire-safety devices (e.g. smoke detectors) in order to support proper maintenance and regulatory compliance. This thesis mainly explores, to great success, the use of artificial intelligence to automatically extracting information from such documents.
Automating this process could reduce time-consuming documentation work and make it easier to keep fire-safety systems up to date—especially for larger facilities such as hospitals, universities, or industrial sites where hundreds of documents may need to be maintained and regularly revised. Although these documents are often easy for humans to interpret at a glance, handling them at a large scale can require extensive manual effort. This would also provide great value for digital platforms seeking to simplify the management of a building's fire-safety system. Our work therefore seek to bridge the gap between traditional documentation and modern digital systems by extracting structured information from complex technical drawings.
We extract this information through a combination of machine learning and algorithmic methods. One important type of document is the so-called service drawing, which resembles a floor plan marked with devices and fire zones. We train a model to detect different kinds of fire-safety devices and use an algorithmic approach for extracting fire zones. Before training the model, we first had to gather and create our own dataset of service drawings from scratch, including the annotation.
In general, our solutions demonstrates high accuracy for most use-cases, albeit with some limitations, where certain visual quirks of the service drawings introduce difficult ambiguities. Overall, the thesis clearly demonstrates that automated interpretation of service drawings is feasible and promising, while also highlighting opportunities for future improvement and development. (Less)
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/student-papers/record/9223713
- author
- Ekstrand, Julius LU and Truong, Victor LU
- supervisor
-
- Kaan Bür LU
- organization
- alternative title
- Datorseendemetoder för att Extrahera Brandsäkerhetsinformation från Serviceritningar
- course
- EITM01 20252
- year
- 2026
- type
- H2 - Master's Degree (Two Years)
- subject
- keywords
- computer vision, service drawings, fire safety, machine learning, artificial intelligence, symbol detection
- report number
- LU/LTH-EIT 2026-1113
- language
- English
- id
- 9223713
- date added to LUP
- 2026-03-18 14:04:05
- date last changed
- 2026-03-18 14:04:05
@misc{9223713,
abstract = {{Fire-safety documentation is an essential part of regular inspections during the building operations and maintenance phase. As help, service drawings exist to detail the positions of various fire-safety devices and their addresses, as well as fire zone coverage throughout the floors and sectors of a building. Automatically extracting this information would be immensely helpful towards, for example, maintaining up-to-date documentation or integration onto digital systems. To that end, this thesis adapts multiple computer vision techniques in order to extract pertinent information: (1) a Keypoint R-CNN model is trained on a custom dataset to detect the symbols of fire devices along with their installation positions, (2) OCR is used to parse addresses and (3) a region merging procedure is implemented to segment fire zones. Additionally, symbols are also assigned to their addresses using the Hungarian algorithm, with Euclidean distance as base cost.
For the symbol detection, we find that crucial model settings had to be changed in order to tailor the model for the domain and achieve satisfactory performance. High accuracy is achieved despite major class-imbalances in the dataset, but the model fails in all cases to detect the true installation positions of devices, whenever these are indicated by an auxiliary line. Assigning symbols and addresses to each other based on distance works well in most cases, while having a few edge cases that are impossible to amend with our method alone, due to inherent quirks in the service drawings. Using region merging for the fire zone segmentation gives us high recall, but low precision. The low precision is mainly due to the oversegmentation step of region merging, and partly due to the existence of hard-to-ignore false positives in most service drawings. Overall, the results across all tasks are promising while still leaving plenty of room for improvement, or alternative approaches, in further work.}},
author = {{Ekstrand, Julius and Truong, Victor}},
language = {{eng}},
note = {{Student Paper}},
title = {{Computer Vision Approaches for Extracting Fire-Safety Information from Service Drawings}},
year = {{2026}},
}